File size: 6,271 Bytes
2fdaed0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
"""Claim 3 (Corollary 26) executed priority + display audit.

Deterministic, offline: parses the cached primary-source documents committed
under evidence/claim_3/sources/ (arXiv abstract pages fetched 2026-07-31 and a
verbatim ar5iv excerpt of the target paper), prints their SHA-256 hashes, and
recomputes every step of the argument with exact rational arithmetic
(fractions.Fraction; no floats anywhere).

Steps executed:
  1. Date audit: extract v1 submission dates of arXiv:2110.13086 and
     arXiv:2312.14141 from the cached pages; both must precede the target's
     2025-09-29 v1 date.
  2. Content audit: both abstracts must describe quantum algorithms for Lasso;
     2312.14141 must use the penalized (ell_1-penalty) form, defeating a
     "different formulation" defense of firstness.
  3. Objective bijection (exact): (1/2)||y-Xb||^2 + t||b||_1 and
     ||y-Xb||^2 + 2t||b||_1 have identical minimizers; verified in closed form
     for the 1-d soft-threshold instance with exact fractions.
  4. Self-citation: the target paper's own related-work text acknowledges
     Chen and de Wolf (2023) "quantum algorithms for Lasso and ridge".
  5. Display counterexample (exact): Corollary 26's printed inequality omits
     lambda in the right-hand minimand; for A=[1], b=[1], lambda=100,
     eps=1/10 the printed guarantee is infeasible with gap 7/40.
  6. Negative controls: a Ridge-only earlier record must NOT falsify Lasso
     firstness; the target compared to itself must NOT count as prior art.
"""

import hashlib
import pathlib
import re
from fractions import Fraction

SRC = pathlib.Path(__file__).resolve().parent.parent / "evidence" / "claim_3" / "sources"
TARGET_V1 = "2025-09-29"


def sha(path):
    return hashlib.sha256(path.read_bytes()).hexdigest()


def parse_abs(path):
    t = path.read_text(errors="replace")
    title = re.search(r"<title>\[([\d.]+)\] (.*?)</title>", t, re.S)
    sub = re.search(r"\[Submitted on ([^\]]+?)(?:\s*\(|\])", t)
    abstract = re.search(r'blockquote class="abstract[^>]*>(.*?)</blockquote>', t, re.S)
    abs_text = re.sub(r"<[^>]+>", " ", abstract.group(1))
    abs_text = re.sub(r"\s+", " ", abs_text).strip()
    return title.group(1), title.group(2).strip(), sub.group(1).strip(), abs_text


MONTHS = {m: i + 1 for i, m in enumerate(
    "Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec".split())}


def iso(date_str):
    d, mon, y = date_str.split()
    return f"{y}-{MONTHS[mon]:02d}-{int(d):02d}"


def soft_threshold_argmin(lam):
    """argmin over x of (x-1)^2 + lam*|x|, exact for lam rational."""
    cand = 1 - lam / 2
    return cand if cand > 0 else Fraction(0)


def halved_soft_threshold_argmin(t):
    """argmin over x of (1/2)(x-1)^2 + t*|x|, derived independently:
    stationarity on x>0 gives x-1+t=0, so argmin = max(0, 1-t)."""
    cand = 1 - t
    return cand if cand > 0 else Fraction(0)


def main():
    print("Claim 3 / Corollary 26 priority + display audit (exact arithmetic)")
    for f in sorted(SRC.iterdir()):
        print(f"  source {f.name} sha256={sha(f)[:16]}...")

    print("Step 1-2: primary-source date and content audit")
    prior = []
    for name in ("arxiv_2110.13086_abs.html", "arxiv_2312.14141_abs.html"):
        arxiv_id, title, sub, abs_text = parse_abs(SRC / name)
        v1 = iso(sub)
        has_lasso = "lasso" in abs_text.lower()
        has_quantum = "quantum" in abs_text.lower()
        penalized = "penalty" in abs_text.lower() or "penalised" in abs_text.lower()
        earlier = v1 < TARGET_V1
        prior.append(earlier and has_lasso and has_quantum)
        print(f"  arXiv:{arxiv_id} v1={v1} earlier_than_target={earlier}")
        print(f"    title: {title}")
        print(
            f"    abstract mentions quantum={has_quantum} lasso={has_lasso} "
            f"penalized-form={penalized}"
        )
    print(f"  prior quantum-Lasso records earlier than target: {sum(prior)} of 2")

    print("Step 3: objective bijection, exact 1-d soft-threshold instance")
    t = Fraction(1, 2)
    m1 = halved_soft_threshold_argmin(t)  # pathwise form (1/2)(x-1)^2 + t|x|
    m2 = soft_threshold_argmin(2 * t)  # target form (x-1)^2 + 2t|x|
    print(
        f"  argmin[(1/2)(x-1)^2+({t})|x|] = {m1}; "
        f"argmin[(x-1)^2+({2*t})|x|] = {m2}; "
        f"equal under lambda_target=2*t: {m1 == m2}"
    )

    print("Step 4: target self-citation of prior quantum Lasso work")
    exc = (SRC / "target_2509.24757_excerpts.txt").read_text()
    quote = "studied by Chen and de Wolf"
    print(f"  excerpt contains '{quote}': {quote in exc}")
    print(f"  excerpt contains 'quantum algorithms for Lasso': "
          f"{'quantum algorithms for Lasso' in exc}")

    print("Step 5: printed-inequality counterexample, exact fractions")
    lam = Fraction(100)
    eps = Fraction(1, 10)
    # left side minimum of (x-1)^2 + 100|x| is at x=0 -> value 1
    x = soft_threshold_argmin(lam)
    left_min = (x - 1) ** 2 + lam * abs(x)
    # right side printed minimand omits lambda: (y-1)^2 + |y|
    y = soft_threshold_argmin(Fraction(1))
    right_min = (y - 1) ** 2 + abs(y)
    bound = (1 + eps) * right_min
    print(f"  min_x[(x-1)^2+100|x|] = {left_min} at x={x}")
    print(f"  (1+eps)*min_y[(y-1)^2+|y|] = {bound} at y={y}")
    print(
        f"  printed guarantee requires {left_min} <= {bound}: "
        f"{left_min <= bound}; impossibility gap = {left_min - bound}"
    )

    print("Step 6: negative controls")
    ridge_only_abs = "quantum ridge regression algorithm for prediction"
    ctrl1 = "lasso" in ridge_only_abs.lower()
    print(
        "  Ridge-only earlier record (Shao 2023, IJMLC 14(1):117-124) "
        f"mentions lasso={ctrl1} -> does not falsify firstness: {not ctrl1}"
    )
    ctrl2 = TARGET_V1 < TARGET_V1
    print(f"  target-vs-itself earlier_than_target={ctrl2} -> not prior art: "
          f"{not ctrl2}")

    verdict = sum(prior) >= 1 and m1 == m2 and left_min > bound
    print(
        "AUDIT RESULT: firstness falsified by >=1 earlier primary source "
        f"(found {sum(prior)}), display inequality falsified exactly: {verdict}"
    )
    fp = f"{sum(prior)}|{m1}|{left_min}|{bound}"
    print(f"RESULTS_SHA256={hashlib.sha256(fp.encode()).hexdigest()}")


if __name__ == "__main__":
    main()